The present invention relates to intra-server network environments for servicing a multitude of hardware controllers over a standard Ethernet LAN (Local Area Network).
A network of computers and Ethernet switches can be redundantly cabled up and configured in order to increase the level of fault-tolerance, for instance, in system control applications. An example of a system control network is described in F. Baitinger et al. “System control structure of the IBM eServer z900”, IBM J. Res. & Dev., Vol. 46, No. 4/5, 2002, pp. 523-535. A similar service network and a solution in support of network configuration are also outlined in the U.S. Pat. No. 7,039,682. In both cases, the network-forming Ethernet switches are not cross-connected via Ethernet.
In this example setup, the redundancy in cabling and configuration is implemented in a way to reduce the number of different parts or configurations needed. Also, the setup must avoid manual configuration steps and intervention by a human being. A part in the setup can be exchanged at anytime without involving reconfiguration.
In one embodiment, the example setup features a physical network structure comprised of two Ethernet switches SWX-A and SWX-B. More generally, the example setup features two classes of computers:
Class (I) computers are called “controllers” herein. Class (I) computers have at least two network interfaces, eth0 and eth1 in terms of Ethernet. The maximum number of Class (I) computers is only limited by the physical connectivity of the Ethernet switches.
Class (II) computers are called “configuration servers” herein. Class (II) computers have only one network interface, eth0 in terms of Ethernet.
The controllers depend on configuration servers for their network configuration (e.g., assignment of Internet (IP) Protocol addresses). The configuration servers could also serve as boot source for the controllers. In this configuration the controllers contain only software with limited capabilities, which is only able to fetch the software code load (e.g. an operating system) from a configuration server.
Configuration servers depend on their location within the physical network infrastructure to discover their own identity, etc. For example, configuration servers can identify themselves as being cabled to switch SWX-A or switch SWX-B, respectively. Depending on its identity, a configuration server must configure its local network interfaces prior to servicing controllers. The configuration servers must be able to configure their network interfaces to be able to serve as a boot source for the controllers. The network configuration for each interface must be unique.
In order to be able to configure a unique IP address, a configuration server must be able to discover its own identity. The uniqueness of its identity serves as a basis for establishing a unique network configuration. If a network environment is redundantly configured, it appears as a perfectly symmetric structure. In order to extract locality information for the controllers, this symmetry needs to be broken.
In the prior art system control network example setup described above, this problem was not addressed. In this environment, the Ethernet switches are part of the computer server hardware itself. However, the problem becomes more severe when using Ethernet switches instead that are part of a normal network infrastructure so as to reduce the costs for the computer server hardware.